About TalkMuch

Why this exists

This started because I am a heavy Joe Rogan listener, five episodes a week. The ones I actually look for are the episodes where the guest dominates the talk time. Those tend to be the better conversations: the host has found someone worth listening to and got out of the way.

The problem is that the only way to find out how much a guest actually got to talk was to listen to the whole thing. A three-hour episode where the host talks over the guest is three hours you do not get back. TalkMuch answers that question before you press play.

What we measure

For every episode we process, we transcribe the audio, separate the speakers, work out which one is the host, strip the sponsor reads, and then measure who held the floor. The number you see is a share of actual conversation time, not of episode length. Advertising breaks and pre-roll are removed before anything is counted.

Alongside the talk split we publish how many questions each side asked, how often the conversation changed hands, and the longest stretch either person spoke without interruption. Those four numbers together tell you more about the shape of an episode than any summary does.

What we do not do

We do not republish transcripts, and we do not rank episodes by quality. A host talking 70% of the time is not doing anything wrong; some shows are built that way. The figures describe a conversation, they do not grade it.

We are also open about where the measurement is soft. Speaker separation gets things wrong, question counts overstate real questions when someone is telling a story, and a few shows defeat our host detection entirely. All of that is written up in the methodology, including the cases where we decided our own tooling was not good enough and said so.

If a figure looks wrong, it might be. Corrections are genuinely welcome, and every one we receive gets checked against the transcript. Get in touch.